Software Engineering Courses (SECourses)•10mo ago

I just wanted to share what I did, using the tips and feedback I received from the people here, can

I just wanted to share what I did, using the tips and feedback I received from the people here, can I ask if there is a proper method of evaluate models

https://www.linkedin.com/posts/italo-thiago-3b3158216_illustriousxl-deeplearning-gpucomputing-activity-7303857523596414976-fgrY?utm_source=social_share_send&utm_medium=android_app&rcm=ACoAADZ7uTcBFffVIJxesBcJt71j4tV36U6H7pk&utm_campaign=copy_link

#illustriousxl #deeplearning #gpucomputing #stablediffusion… | Ital...

Just finished a preliminary training run with Illustrious XL using localy 2xRTX 3090s. 36 hours, 2 epochs, and 1344 CC0 images, Here some exciting initial results. The speed and quality are impressive! More in-depth analysis to follow. #IllustriousXL #DeepLearning #GPUComputing #StableDiffusion #DomesticCluster #AIart #GrapichDesign #Genai

PPixel.dust I just wanted to share what I did, using the tips and feedback I received from t...

JotaTerrasa•3/8/25, 8:48 AM

Great work!

JotaTerrasa•3/8/25, 8:49 AM

Can you share weights?

FFurkan Gözükara SECourses sadly thinlinc client is not very good. <@1167168099838074940> <@347893716528136...

Timson•3/8/25, 9:37 AM

I use Syncthing - it's free, multi-platform, don't have much issues with large files

Timson•3/8/25, 9:38 AM

I include xml config with my server initialisation script, so it requires just 2 clicks per worker to set up

TTimson I include xml config with my server initialisation script, so it requires just 2...

Furkan Gözükara SECourses•3/8/25, 10:51 AM

nice

JJotaTerrasa Great work!

Pixel.dustOP•3/8/25, 2:36 PM

Yes Sooner

FFurkan Gözükara SECourses epoch doesnt change according to batch size

spt999•3/8/25, 4:17 PM

sorry for this question was tired from attempts

Now done x2 upscale photos, much better results, thank you!
Try today 20-80 epochs, decided add 40 epochss to compare. Very good results

Sspt999 sorry for this question was tired from attempts 🤣 Now done x2 upscale photos...

Furkan Gözükara SECourses•3/8/25, 5:09 PM

great

Sspt999 sorry for this question was tired from attempts 🤣 Now done x2 upscale photos...

AiInfluence•3/8/25, 7:25 PM

what upscaler did you use ? Supir?

TrumpYouAll•3/9/25, 10:47 PM

@Furkan Gözükara SECourses Would you please update the file of Koyha (https://www.patreon.com/posts/click-to-open-post-used-in-tutorial-112099700) (File name: Kohya_FLUX_DreamBooth_v17.zip)? I would like to try using 5090 to use Kohya (Windows) to train Flux Dreambooth

Thanks!

Patreon

Kohya FLUX Fine Tuning (Full Checkpoints) Training Full Tutorial Fo...

Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon

TTrumpYouAll <@205854764540362752> Would you please update the file of Koyha (https://www.pa...

Furkan Gözükara SECourses•3/9/25, 10:58 PM

just updated for you

Furkan Gözükara SECourses•3/9/25, 10:58 PM

download v18 and read change logs

FFurkan Gözükara SECourses download v18 and read change logs

TrumpYouAll•3/9/25, 11:00 PM

First, LOVE YOU !
Second, I have been refreshing Massed Compute for GPU for another training

besides 5090 training that now I am going to use.
So many people using A6000 at the moment!

TTrumpYouAll First, LOVE YOU ! Second, I have been refreshing Massed Compute for GPU for ano...

Furkan Gözükara SECourses•3/9/25, 11:00 PM

yes @computersheep said they gonna add more hopefully soon

Furkan Gözükara SECourses•3/9/25, 11:02 PM

by the way you can use 48 gb config on 32 gb

FFurkan Gözükara SECourses by the way you can use 48 gb config on 32 gb 😄

TrumpYouAll•3/9/25, 11:03 PM

@Furkan Gözükara SECourses U know, I am not really that knowledgeable in this area like most of you here. I normally just use the EXACT config and steps you showed in your YouTube.

TTrumpYouAll <@205854764540362752> U know, I am not really that knowledgeable in this area li...

Furkan Gözükara SECourses•3/9/25, 11:07 PM

sure it will work no worries

FFurkan Gözükara SECourses sure it will work no worries

TrumpYouAll•3/9/25, 11:08 PM

1 thing about how good your tutorials are is that they are so clear that it is dumb-proof. To be honest, I got zero clue what is going on with all the names in the SWAMUI and Koyha. U have made it very clear.

TTrumpYouAll 1 thing about how good your tutorials are is that they are so clear that it is d...

Furkan Gözükara SECourses•3/9/25, 11:08 PM

thanks a lot. some people complains but someone has to make this way

Furkan Gözükara SECourses•3/9/25, 11:09 PM

i got to sleep hopefully see you tomorrow

TTrumpYouAll 1 thing about how good your tutorials are is that they are so clear that it is d...

Furkan Gözükara SECourses•3/9/25, 11:19 PM

Don't forget to change xformers to SDPA. So easy look xformers on gradio interface

FFurkan Gözükara SECourses Don't forget to change xformers to SDPA. So easy look xformers on gradio interfa...

TrumpYouAll•3/9/25, 11:21 PM

Even though i got no idea what u talk about, I will definitely look into the parameters (?) section to change it. I don't 'forget'. I just don't know

FFurkan Gözükara SECourses by the way you can use 48 gb config on 32 gb 😄

TrumpYouAll•3/10/25, 12:15 AM

@Furkan Gözükara SECourses Please kindly or u tell me which parameter I need to change, it turns out to be 440,000 steps for me to use 48GB config. (I guess something wrong with the parameter)
Thanks

TTrumpYouAll <@205854764540362752> Please kindly or u tell me which parameter I need to chang...

Furkan Gözükara SECourses•3/10/25, 10:00 AM

it is folder repeat

Furkan Gözükara SECourses•3/10/25, 10:00 AM

fix your folder repeat to 1

RyanBelnap•3/10/25, 11:54 PM

I'm facing a challenge with DreamBooth training and sourcing high-quality images for training. While web scraping tools are a potential solution, they often retrieve compressed or resized images, which reduces the quality needed for optimal training results.

Does anyone have suggestions or alternative to obtain large amounts of high-resolution images without compression?

RRyanBelnap I'm facing a challenge with DreamBooth training and sourcing high-quality images...

Furkan Gözükara SECourses•3/11/25, 12:01 AM

yes unsplash

Furkan Gözükara SECourses•3/11/25, 12:01 AM

or pexel

Furkan Gözükara SECourses•3/11/25, 12:01 AM

directly crawling

RyanBelnap•3/11/25, 12:10 AM

Thanks, Dr. Furkan! I appreciate the suggestion. The web scraper I’m currently developing is designed to handle this task manually, but before I dive into a long weekend of coding, I’d like to explore whether there’s an API for Pexels or if anyone has alternative suggestions. Since this is my first time developing an app, I’m open to insights that could help streamline the process. Let me know if you have any recommendations!

RRyanBelnap Thanks, Dr. Furkan! I appreciate the suggestion. The web scraper I’m currently d...

Furkan Gözükara SECourses•3/11/25, 12:12 AM

maybe you can find open source crawlers

Furkan Gözükara SECourses•3/11/25, 12:12 AM

they also have api

Furkan Gözükara SECourses•3/11/25, 12:12 AM

i mean at least unsplash

RyanBelnap•3/11/25, 12:16 AM

looking into it now, maybe this will work! Thankyou for the help and videos my friend.

AAiInfluence what upscaler did you use ? Supir?

spt999•3/11/25, 11:21 AM

yes

ikros•3/11/25, 12:56 PM

Hi guys I have been having some issues with the fine tuning of flux.

I have been trying to fine tune a car with 39 images and I am having poor results with the configuration 48GB_GPU_28200MB_6.3_second_it_Tier_1.json and with 145 epochs.

The car looks good but not that good and also the main problem I have been having is that whenever i just divert the prompt from "a picture of a sks car" then the model just reproduces a random car.

Also in the training data i don't have any prompt for my 39 images. Do you think the problem might stand there?

Original message was deleted

Furkan Gözükara SECourses•3/11/25, 2:44 PM

it bleeds : https://medium.com/@furkangozukara/thoroughly-experimented-with-fine-tuning-dreambooth-training-of-flux-dev-de-distill-pixelwave-13fab8c625bf

Medium

Thoroughly experimented with Fine-Tuning / DreamBooth training of F...

Thoroughly experimented with Fine-Tuning / DreamBooth training of Flux-dev-de-distill, PixelWave v03, Verus Vision and base FLUX Dev model…

Iikros Hi guys I have been having some issues with the fine tuning of flux. I have bee...

Furkan Gözükara SECourses•3/11/25, 2:45 PM

use ohwx car

Furkan Gözükara SECourses•3/11/25, 2:45 PM

try to describe car in prompt after trainig

Furkan Gözükara SECourses•3/11/25, 2:45 PM

also if still bad then you need a better dataset

Furkan Gözükara SECourses•3/11/25, 2:45 PM

also i recommend latent 2x upgrade

Furkan Gözükara SECourses•3/11/25, 2:45 PM

tips-info

ikros•3/11/25, 2:57 PM

Thank you! I will try that

sai_150•3/13/25, 3:29 PM

Hi all, I'm following the youtube video to train Lora using kohya ss in kaggle. I'm getting this error when run code the launch gui. ERROR: Exception in ASGI application. Is anyone able to find a fix to this issue? Please let me know if I'm asking in the wrong channel.

Ssai_150 Hi all, I'm following the youtube video to train Lora using kohya ss in kaggle. ...

Furkan Gözükara SECourses•3/13/25, 3:33 PM

hi which zip file you are using or kaggle file?

Furkan Gözükara SECourses•3/13/25, 3:33 PM

can you give me link

sai_150•3/13/25, 3:34 PM

I'm following the commands in this page. https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/How-To-Do-Free-Google-Colab-Like-Kaggle-SDXL-LoRA-Training.md

GitHub

Stable-Diffusion/Tutorials/How-To-Do-Free-Google-Colab-Like-Kaggle-...

FLUX, Stable Diffusion, SDXL, SD3, LoRA, Fine Tuning, DreamBooth, Training, Automatic1111, Forge WebUI, SwarmUI, DeepFake, TTS, Animation, Text To Video, Tutorials, Guides, Lectures, Courses, Comfy...

sai_150•3/13/25, 3:35 PM

Ssai_150 I'm following the commands in this page. https://github.com/FurkanGozukara/Stabl...

Furkan Gözükara SECourses•3/13/25, 3:38 PM

are you using v22 ?_

Furkan Gözükara SECourses•3/13/25, 3:38 PM

https://www.patreon.com/posts/kohya-sdxl-lora-88397937

Patreon

A Kaggle Notebook to Train Stable Diffusion 1.5 and XL (SDXL) on a ...

Get more from SECourses: FLUX, Tutorials, Guides, Resources, Training, Scripts on Patreon

I just wanted to share what I did, using the tips and feedback I received from the people here, can

Similar Threads

Similar Threads

Similar Threads